Download Preliminary Considerations for Questionnaire

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Psychometrics wikipedia , lookup

Taylor's law wikipedia , lookup

History of statistics wikipedia , lookup

Bootstrapping (statistics) wikipedia , lookup

Resampling (statistics) wikipedia , lookup

Gibbs sampling wikipedia , lookup

Sampling (statistics) wikipedia , lookup

Misuse of statistics wikipedia , lookup

Transcript
RESEARCH METHODS IN BUSINESS
STUDY MATERIAL SUPPLIED FOR THE COURSE BM 512 OF MASTER OF BUSINESS ADMINISTRATION
UNDER THE SCHOOL OF MANAGEMENT SCIENCES , TEZPUR UNIVERSITY, and prepared by
MRINMOY K SARMA. It should be noted that the material is supplied as a part of the course curriculum and is in
no way exhaustive.
Business Research has been used as an instrument for reducing managerial errors in decision
making. Some of the researches conducted in the actual business field are routine, while some are
commissioned for a specific one-time purpose. However, it may not be taken for granted that research
will always provide with the right answer to a particular problem. In fact, it is reported that almost 70% of
the researches conducted all over the world either offers inaccurate information or misleads the
decision maker. When Xerox initially wanted to launch the copier machine, out of three market
researches commissioned, two advised the company against the proposed launch. The remaining
agency predicted a turnover of only 8000 machines within the next 6 years. Xerox launched the product
and sold more than 80000 (eighty thousand) pieces within the next 3 years. New Coke, Ford Edsel are
few of the classic examples of the victims of market research. There are many reasons for which a
research may lead to inaccurate results. Errors may creep in at every stage of the research process.
For example, sometimes a wrong approach might be taken, while in some other cases a wrong
instrument may be adopted. Therefore, adherence to the correct method of conducting a research is
the prerequisite of success.
The Research Process:
The following is the flow of research process. However, minor variations are found in different
books.
Establish the need for information
Specify research objectives and information needs
Determine research design and source of data
Develop the data collection procedures
Design the sample
Collect the data
Process the data
Analyse the data
Present research results
We will be discussing here the highlighted portions of the process.
Designing the Sample:
This includes the following jobs:
01.
Decide about the method of sampling to be employed
02.
Decide about the sample size
03.
Select a sample
Research Methods in Business
Class Notes: Tezpur University MBA Programme
2
© Mrinmoy K Sarma
Before proceeding let us have a look at the cases where sample drawing is necessary. In many
a situations it is not possible to make a census (100% enumeration of the total elements), like in case of
measuring the attitude of the target customers on a new packaging or on a new advertising campaign.
For that matter, in case of Opinion Poll on the results of a general election it is just next to impossible to
contact all eligible voters and to know their preferences. The cost involves in doing so would be too
much (in fact, as much as the cost of the general election itself) which would negate the usefulness of
the Opinion Poll. Thus, we do take recourse to sampling methods for the following reasons:
Economy in Money and time,
Not to destroy or contaminate the population, and
Sometimes for more accuracy. (explanation will be offered in the class)
At this juncture the meaning of the following frequently used terms are explained.
Elements:
About which information is sought; may be human, Product, Stores, Company
etc.
Population (or Universe): Aggregate of all the elements defined prior to selection of the
samples. Population must be defined in terms of
Elements :
…as defined above
Sampling Units: The element(s) which are available for selection at some stage(s) of
the sampling process, like Chemical Engineers may be sampled from
a business organisation whose turnover is more than Rs.5 crore in the
last financial year.
For example, if we are to select samples from among the males aged
over 50 years form households from blocks of the cities having more
than 5 lac of population we have the following multistage Sampling
Units:
Primary S.U.: Cities above 5 lac of population
Secondary S.U.: City Blocks
Tertiary S.U.: Households
Final S.U.: Males aged over 50 years.  this is the element
Extent:
Coverage in respect of geographical area of the elements or sampling
units.
Time:
The time period within which samples are drawn.
Sampling Frame
In case we are interested in drawing samples through probabilistic methods (which are
explained later in this material), a sampling frame is necessary. A sampling frame is the means of
representing the elements of the population. This may be a Telephone Directory, Employee Register or
a Voter’s list. Though in many social science research problems the sampling frame often difficult to
define, proper car must be taken to find and establish a sampling frame before proceeding further with
the research process. Maps also serve frequently as sampling frames. This is useful in case of area
sampling.
A perfect sampling frame is one in which every element of the population is
represented once and only once. Examples of perfect frames are rare, however, specially when we are
interested in sampling from any appreciable segment of a human population.
Errors in the sampling frame may be exclusion or multi inclusion of elements in the
frame. These are known as frame errors.
Research Methods in Business
Class Notes: Tezpur University MBA Programme
3
© Mrinmoy K Sarma
However, one does not need a sampling frame to take a non-probability sample.
Sampling Unit
A sampling unit is the basic unit containing the elements of the population to be
sampled. It may be the element itself or a unit in which the element is contained.
For example, if we want to sample males of 21 years of age, it might be possible to
meet each sample directly. In this case, the sampling unit would be identical with the element. However,
if we were interested in sampling children below the age of 10, it would be easier to meet them in their
residence in presence of their parents. In this case, the sampling unit is the household and the element
is the child below 10 years of age.
In any case of interview, further specification of the sampling unit is required like should
we interview the person who answers the doorbell first, if he/she is an element of our study population.
Interviewing whoever remains present at home, may sometime lead to overrepresentation of women
and elderly persons in the sample. Therefore, for surveys, where random samples of an adult
population is desired, a random selection must be made from the adult residents of each household
(sampling unit in this case). The ‘next birthday’ method (where the birthdays of adults are taken and the
person whose birthday falls first is interviewed) is the simplest method to use in such a situation. Many
innovative methods can also be used, which might be the result of creativity of the researcher.
We can select sampling units in different stages. Suppose we are interested in
selecting samples which are residents of a town having a population of 1 lac or more, having residing
near a main street, and female more than 30 years of age, we will have to take up sampling units as
follows:
Primary s.u
:
Cities with more than 1 lac population.
Secondary s.u :
Main streets.
Tartiary s.u
:
Households.
Final s.u
:
Females of more than 30 years of age.
Statistic:
These are the characteristics of a sample.
Parameter:
These are the characteristics of a population.
Symbols:
for
Population
for
Samples
Size:
N
n
Mean:

x
S.D.:

s
Research Methods in Business
Class Notes: Tezpur University MBA Programme
4
© Mrinmoy K Sarma
Steps in Selecting a Sample:
Define the population in terms of
1. Elements
2. Units
3. Extent and
4. Time
Identify the sampling frame
Determine the sample size
Select a sampling procedure
These two steps can also be
performed simultaneously
Select the sample
We have already discussed about the first two steps. The third step would be discussed later,
along with various methods of drawing sample.
There are many different sampling procedures by which researchers may select their samples.
But one fundamental concept must be dealt with at the outset - the distinction between probability and
non-probability sampling.
In probability sampling each element of the population has a known chance of being selected
as a sample. The sampling is done by mathematical decision rules that leave no discretion to the
researchers. It is to be noted that there is a difference between known chance and equal chance. Equal
chance probability sampling is only a special case, which is called simple random sampling. Probability
sampling gives us a distinct advantage over non-probabilistic sampling; that is, this allows to calculate
the likely extent to which the sample value differs from the population value of interest. This difference
is called sampling error.
In non-probability sampling, the selection is based on some part of judgment of the
researchers. There is no known chance of any particular element in the population being selected.
Therefore, it is not possible to calculate the sampling error.
The different available fundamental sampling procedures are,
1.
2.
3.
Probability Sampling
Non-probability sampling
Simple random Sampling
Stratified Sampling
Cluster Sampling
a.
Systematic Sampling
b.
Area Sampling
1.
2.
3.
Convenience Sampling
Judgment Sampling
Quota Sampling
Convenience Sampling:
This is based on the convenience of the researchers. Therefore, it is unclear about the actual
population. One such example would be people-on-the-street interview by a television interviewer.
Research Methods in Business
Class Notes: Tezpur University MBA Programme
5
© Mrinmoy K Sarma
In this case, the difference between the population value of interest and the sample value is
unknown, in terms of both size and direction. And we can not measure the sampling error. And clearly
therefore, we can not make any definite conclusive statement about the result of such sampling. Thus,
this type of sampling is appropriate at the exploratory stage of the research.
Judgment Sampling (or Purposive Sampling):
Here the basis is expert opinion on the usefulness of selecting a particular element as the
sample. For example, in test marketing, a judgment is made as to which cities would constitute the best
ones for a particular product targeted to a particular group of customer. Or the decision about
interviewing a particular dealer regarding a new incentive scheme would definitely call for expert opinion
or past experience of the researchers.
Here also, the degree and direction of error is unknown, and definite statement regarding
findings of the survey is not meaningful. However, if the expert’s judgment was valid, then such result
would be more representative than that of a convenience sampling.
Quota Sampling:
This is a special type of purposive sample. Here the researcher takes explicit steps to obtain a
sample that is similar to the population on some “pre-specified” controlled characteristics. e.g., a
interviewer may be instructed to select half of the interviewees from people 30 years of age and older
and the other half under the age of 30. Here the control characteristic is age of the respondents. In real
life the interviewer may have to face more control characteristics like age, educational background,
place of residence
(urban or rural), etc. In such a situation the researcher will have to use his
discretion to obtain samples from each of the categories equally, as far as possible.
This method is the most widely used among all non-probabilistic methods of sampling. Almost
47% of the American firms use it frequently and almost 39% use this method “sometimes”.
In the above example, if the age is divided in four categories ( under 15, 16 to 25, 26 to 50 , and
51 and above) place of residence into two categories, educational level into four categories ( under
HSLC, Graduates, Post Graduates and Professional degree holders) and add another variable ,income
level ,with five categories ( income per month below (Rs.) 5000/-, 5001/- to 7500/-, 7501/- to 10000/and 10001/- and above) we will have 4 x 2 x 4 x 5 = 160 sampling cells. In this case we may have
equal number of representation from each of the 160 cells. The number of representation can be
derived by the following method.
Multiply the population size by desired proportion from each cell
Disadvantages of quota sampling are - the proportion of respondents assigned to each cell
must be accurate and up-to-date. This is often difficult and impossible. The “proper” control
characteristics must the selected . It is also not possible to include more variables due to practical
difficulty. Even these problems are solved, interviewers may not be able to select actual respondents for
interviews.
This method is useful in preliminary stage of the research , if done with care, they can provide
more definite answers. However, such results are less valid than a probability sampling.
Simple Random Sampling:
This is the most frequently used method of sampling. In such sampling we select the sample
randomly. But for this method to be successful two preliminary conditions must be fulfilled:
1.
Each element must get an equal chance of being selected
Research Methods in Business
Class Notes: Tezpur University MBA Programme
2.
6
© Mrinmoy K Sarma
Each combination of the n sampling elements has an equal chance of being selected.
The first condition says that every element must be equally likely top be selected, like a
blindfolded man taking out a ball from a bag full of balls of different colours and of equal size and
weight.
The second condition says that if we are to select 4 samples from a population of 16 elements,
than every combination (there would be 16C4 nos. of combinations) would have equal chance of being
selected.
For simple random sampling we use Random Number Table (supplied in the class).
For calculation of population parameter following formulae are used.
=
X
2=
 x   
N
=
OR
N
and
 fX
f
S.D.=
2
for drawn samples:
Sample mean x 
x
n
(X)2
(
Sample Variance s =  X 2
)/ df
_
n
df = n- ( no. of statistic calculated, generally 1)
S.D. of the sampling distribution (the Standard Error)
Sx=
s
n
The Classical Theory of Statistics:
This theory would help the students in understanding the properties of normal distribution which
is very important in understanding the behaviour of sampling distribution and useful in grasping the
hypothesis testing fundamentals.
The sampling distribution of mean and statistical inferences:
In any population there are many possible sample groups. The classical statistical inference is
based on what happens when one repeatedly selects different sample groups from the population.
If we repeat calculation of mean of the sample twice, thrice, and n times we find that the sample
mean closer in value to the population mean , would tend to repeat more frequently than others. If, now,
we plot this mean value in a graph we would find a bell shaped curve. This is normal distribution curve.
Research Methods in Business
Class Notes: Tezpur University MBA Programme
7
© Mrinmoy K Sarma
This distribution is known as sample mean distribution or sample distribution . It is important in two
ways:
1.
The sample mean in this distribution is distributed around the population mean in known way,
2.
Using this distribution, we can determine how closely the sample statistics are distributed
around the population parameter.
FOR FORMALISATION OF THE NATURE OF SAMPLING DISTRIBUTION OF THE MEAN we
consider the Central Limit Theorem of Statistics:
1.
If a population distribution of a measure is normal the sampling distribution of the mean is also
normal.
2.
If the population distribution is non-normal the sampling distribution of the mean approaches
normal as the sample size increases.
3.
The mean of sampling distribution of the mean is population mean. In the type of situation in
which the expected value of the mean of the sampling distribution for the statistics is the parameter or
population value the statistic is said to be unbiased.
4.
The S.D. of the sampling distribution of the mean is the population S.D. divided by the square
root of the sample size. This value is often called the standard error of the mean.
As in practice we do not know  or  so we estimate them with X and s
68%
95%
99.7%
There is an important aspect about the normal curve.
1. 68% of the cases will be within + 1 standard deviation of the mean
2. 95% of the cases will be within + 2 standard deviation of the mean
3. 99.7% will be within + 3 standard deviation of the mean.
The above diagram depicts a normal curve with area under +1,+2, +3 standard deviation of the
mean.
Other characteristics of the normal curve:
a.
b.
c.
d.
The curve is of a single peak; it has the bell shape.
The mean ( ) lies at the centre of the normal curve
Median and mode are also at the centre, i.e., mean = mode = median
Two tails never meet the horizontal axis.
CERTAIN FREQUENTLY USED TERMS IN RANDOM SAMPLING:
Confidence interval:
Let us assume the followings for ease of our calculation and understanding:
s = 2.88 and n = 5 as found in particular sampling statistics calculation.
Research Methods in Business
Class Notes: Tezpur University MBA Programme
8
© Mrinmoy K Sarma
and mean of the sampling distribution is 22.6. ( already calculated)
Then the standard error or the (s.d. of the sampling distribution)2 is
Sx

2.88
5
 1.3
Now let us calculate the size of the intervals at + 1 standard deviation from the mean, +2 s.d.,
and + 3 s.d. from the mean. At +1 s.d. the inteval is
22.6+ 1.3 = range 21.3 and 23.9
We know that 68% of the means from our sampling distribution are contained in this interval if
our calculated sample mean X is truly the mean of the sampling distribution. Thus at the + 2 s.d. the
interval is
22.6 + 2 ( 1.3) = 22.6 + 2.6 = range 20.0 and 25.2
Again we know that 95% of the means from our sampling distribution are contained in this
interval if our calculated sample mean X is truly the mean of the sampling distribution.
These intervals are known as Confidence intervals. The first interval was 68%, the second one
was 95% confidence interval.
Sampling fraction =
n
N
The sampling fraction can be used to estimate the total population usage of product or services
from the total sample usage. Suppose that a sample of 5 students out of the population of 50 ) used a
total of 35 liters of petrol per week . Then the estimated total population usage of petrol would be,
Total sample usage
Sampling fraction
= 35/ (5/50) = 35/.1 = 350 liters.
Determining the sample size:
After understanding of sampling error and non-sampling error, let us have a look at the sample
size determination.
In simple random sampling for a known sample size , we calculated the confidence interval of
our estimate at a given level of confidence. To do this for a continuous measure we have the following
information:
1.
An estimate of the mean, x
2
An estimate of the standard deviation, s
3
A sample size
4
A level of confidence
5
Using items two and three, we calculated the standard error, s x. We then calculated the relevant
confidence interval. The equation to do this at 95% confidence level was
Research Methods in Business
Class Notes: Tezpur University MBA Programme
Confidence interval =
x 2
9
© Mrinmoy K Sarma
s
n
We have calculated the x and s, and we know n, so we can solve this equation for the
confidence interval. Or, we could calculate the precision we obtained using part of the above equation
as follows,
Precision =
s

n
Now, suppose we want to reach a level of given precision. If we have a value for s, we can
solve this equation for the required sample size.
Let us illustrate this. Suppose at the 99.7 % level of confidence we wish to obtain an estimate of
the mean age of a target segment for a new magazine that is within + 1.5 years of the true mean age.
In addition , we will assume that we have an estimate of s= 6.0. The required sample size is obtained by
solving the following equation for n:
Precision =
+ 1.5 years

=+3
1.5
=
s
n
6
n
18
n

1.5n
n
=
=
18
12
n
=
144
this 3 comes from the equation of
determining the area under the
normal curve at.99.7% confidence
level.
Thus, with a sample size of 144 will give us a precision of + 1.5 years, if s= 6.
Here we have used absolute precision. If we express the precision level in terms of percentage
we call it relative precision. In this situation the formula would be
b.
x
=+
s
n
b = percentage of precision level
x = the estimate of mean
s
Here also the
should be multiplied by the coefficient at desired level of coefficient.
n
Research Methods in Business
Class Notes: Tezpur University MBA Programme
10
© Mrinmoy K Sarma
The most disturbing thing in our calculation of sample size is that we need to know the value of
s for absolute precision and a value of
s
in case of relative precision.
x
If we have these values, in all likelihood, we already know what we intend to know. Also, for
absolute precision the required sample size varies (a) inversely with the size of the precision desired,
(b) directly with s, and (c) directly with the size of the confidence level desired. In most studies we want
to measure many variables. To the extent that they differ in terms of precision desired , s, or confidence
level, the required sample size will differ. There is no one-sample size that is statistically optimal for any
study. The only way to assure the required precision would be to select the largest sample.
However, if a researcher has experience with the problem at hand , then very accurate estimate
of s are likely to be available at the time the sample size is being planned.
No one should accept the sample size generated by the statistical formula blindly. One reason
for not doing so is the existence of non-sampling errors. Non-sampling errors increase as the sample
size increases. Therefore, a carefully done study of 200 samples might give, sometimes (but not
always), better result than that from a sample size of 2000.
Sample size and other factors:
In any kind of business research one is always to find out a compromise between technical
elegance and practical constrains. These constrains are definitely effect the decision regarding the
sample size . Some of them are 1. Study objectives
2. Time constrain
3. Cost constrain
4. Audience acceptability
5. Data analysis procedure
The sample size determination procedure becomes complicated with more number of variables
are taken for analysis.
Other methods of Sample size determination:
1. Unaided judgement
2. All-you-can-afford
3. Average size in similar studies
4. Required size per cell in case of quota sampling
5. Traditional statistical methods ( as discussed above)
Stratified Sampling:
Stratified sampling calls for division of the total population into some sub groups and then
collect samples from each such group. The underlying objective of this method is to reduce the
standard error of the estimator. Thus the confidence interval we calculate will be smaller.
The method:
1.
2.
Divide the total population into mutually exclusive and collectively exhaustive
groups or strata.
Perform an independent simple random sample in each stratum.
Let us have the following notations:
Research Methods in Business
Class Notes: Tezpur University MBA Programme
Nst.1
Nst.2
nst.1
nst.2
Xst.1
Xst.2
s2st.1
s2st.2
11
© Mrinmoy K Sarma
= population in the stratum 1
= population in the stratum 2
= sample size in stratum 1
= sample size in stratum 2
= sample mean of the stratum 1
= Sample mean of the stratum 2
=sample variance of the stratum 1
= sample variance of the stratum 2
Disproportionate Stratified Sampling:
The overall sample size n be allocated to strata on a disproportionate basis with the population
sizes of the strata. Normally, we can reduce the standard error by sampling heavily in strata with higher
variability. Therefore, to reduce standard error we should draw heavy sample from strata whose
population is more variable in nature. Past experience and earlier studies can give us data regarding
variability in strata. Therefore, in such situation the samples from each strata differs in number and this
is known as disproportionate stratified sampling.
Cluster Sampling:
In cluster sampling a cluster or a group of elements are randomly selected at one time unlike
the other methods, where individual elements were picked up one by one. Therefore, the population
here also, should be divided into mutually exclusive and collectively exhaustive groups. We can select
randomly any of these groups. If we select groups thus, and use all the elements in the selected groups
as samples, it is known as one-stage cluster sampling. Or if we had selected a random sample
elements from within the selected groups this is known as two-stage cluster sampling.
In cluster sampling we try to form groups as heterogeneous as in the population (this is just
opposite to the stratified sampling, where, we try to formulate strata as homogeneous as possible), so
that selected samples from any of the groups would be representative of the total population as a
whole. If the groups are less heterogeneous than that of the population, then the standard error from
such sampling will be more than that of the simple random sampling.
Systematic Sampling:
In systematic sampling the researcher select every Kth element in the sampling frame , after a
random start somewhere within the first k elements. Suppose we want to select a systematic sample of
n=5 from a population size of 50 then the K will be
k= N/n = 50/5 = 10
The steps will be,
1. Obtain a random number between 1 and 10, This element will be picked up first.
2. Add 10 to this random number. This element will be the second element of the sample. Then add
another 10 and pick up that sample and so on.
Systematic sampling is easy and cheap to use. This is a close substitute of the simple random
sampling. Here, we do not need the complete sampling frame unlike in the simple random sampling.
However, the problem of periodicity might occur without the knowledge of the interviewer.
Area Sampling:
In each of the above sampling procedure a complete accurate listing of the elements of the
population is required . Unfortunately , for a great many business research applications such lists are
Research Methods in Business
Class Notes: Tezpur University MBA Programme
12
© Mrinmoy K Sarma
impossible to find. Therefore, in area sampling the area where the sample reside is taken into
consideration and samples are selected accordingly. Frequently used version of are sampling is Multi
Stage Area Sampling. The steps involved in MSAS are described below. Here a hypothetical case of
MSAS is described.
Stage 1:
Divide India in 5 zones. North, East, South, West and Northeast.
Stage 2:
A listing is made of the states fall within each of the zones:
Thus in the northeast seven states namely, Assam, A.P., Mizoram, Manipur, Nagaland,,Tripura
and Meghalaya. This type of listing is done for every zones.
Then a state from the selected zone in stage 1 is selected.
Stage3:
Another list of major cities is to be prepared of the selected state. Thus if Assam was selected
randomly the major cities will be, Guwahati, Tezpur, Jorhat, Tinsukia, Nagaon, Nalbari , Silchar,
Dibrugarh, Mangaldai ( or may be all district Hqs). Then a particular city is selected
Stage 4:
Then a ward of that particular city is selected .
Stage 5:
Then a street of that particular ward
Stage 6:
Then a house from the selected street.
Research Methods in Business
Class Notes: Tezpur University MBA Programme
13
© Mrinmoy K Sarma
Concepts of Measurement and Scaling
Once a research problem is defined and a particular plan of action is chosen to solve the
problem, three main components needed to be decided. Designing questionnaire, determining sample
size, and sampling technique including finalisation of field procedures. In many cases of practical
interest -- new product concept testing, corporate image measurement, advertisement effectiveness
measurement, and the like the researcher will seek information of various nature -- even of
psychological dimensions. If useful data are to be obtained from the field the researcher will have to
exercise cautions in deciding what is to be measured and how to make the measurement. This is to be
decided before preparing the questionnaire and starting field operations. Thus measurement of a
business phenomenon is fundamental to providing any meaningful information for decision making. The
objective of measurement is to transform the characteristics of objects into a form that can be analysed
later. Therefore, the object under study must be defined so that this can be measured. Many
researchers (specially the student researchers) blunder here and try to measure the object without pre
defining the objects under study.
Definitions in Research: An important part of practice of research entails construction, use, and
modification of definitions of objects. ”Attitude", "aggressiveness" "leadership capabilities" "job
satisfaction" can not be measured as it is, since these things can not be described objectively. Vague
definitions can not be measured, and therefore, cannot be used for further decision making.
Definitions can be distinguished into two classes --Constitutive definition and operational
definition. Constitutive definition is roughly similar to a dictionary definition. An operational definition
establishes the meaning of an object through specifying what is to be observed and how the
observation is to be made. Measurements and operational definition go together. That means the
researcher will have to get an operational definition of the object under investigation. Let us take the
example of "job satisfaction". Operational definition would suggest the individual components to be
measured to collectively arrived at a decision regarding "job satisfaction" of a particular group of
workers.
In business research measurement process involves using numbers to represent the business
phenomena under investigation. Stated formally, the empirical system includes marketing phenomena,
such as buyer reaction to products or advertisement, while the abstract system includes the numbers
used to represent the business phenomena.
The Measurement Process
Empirical system
Physical Sciences
Measurement
Social Sciences
Abstract System
Number System
Measurement Defined: It is the assignment of numbers to characteristics of objects or events
according to rules. Effective measurement is possible when the relationship existing among the objects
or events in the empirical system directly correspond to the rules of the number system. If this
correspondent is misrepresented, measurement error occurs.
Types of Scales:
Scales have been classified in terms of the four characteristics of the number system. These
scales of measurement are nominal, ordinal, interval and ratio. The following chart compares the four
scales of measurement. The understanding of the types of scales of measurement is necessary
because analytical procedure differs according to the type of the scale.
Research Methods in Business
Class Notes: Tezpur University MBA Programme
14
© Mrinmoy K Sarma
Characteristics of Measurement Scales
Scale
1.
Nominal
2.
Ordinal
3.
Interval
4.
Ratio
Number system
Unique definition of numerals
( 0, 1,….,9)
Order of numerals
1<2, 2<3, 5>4 etc.
Equality of difference
3-2=8-7
Equality of ratios
2/4 = 4/8
Characteristics
No origin, no order and no
equality of differences
No origin, no equality of
difference. But with order
No origin. But with order and
equality of difference
With origin, equality of ratios,
and order
Questionnaire
A questionnaire is a formalised schedule for data collection. The basic job of questionnaire is to
measure the variables under study. This is the most widely used primary data collection techniques,
and used throughout the world for social science researches.
Questionnaire is one of the methods through which the empirical system definition (operational
definition of the variable) is converted into abstract system of the number system, which measures the
variables under consideration. If something goes wrong in this process measurement error occurs.
The next paragraphs are used to discuss the principles of perfect questionnaire construction,
which can reduce this error to the minimum.
There are 5 basic components of the questionnaire. They are (1) Identification Data, (2)
Request for Cooperation, (3) Instruction, (4) Information Sought, (5) Classification Data.
Identification Data contain the respondent’s name, address and other relevant information,
which might be used at a later date to identify the
respondent from the questionnaire. Often these information
First Page of Questionnaire
are collected from some secondary sources like the sampling
Identification
Data
frame before selection of sample and therefore, as one of the
first steps of data collection. Normally, this information is
written in a separate sheet of paper, which is not shown to
Request for Cooperation
the respondents. Many researchers put a code mark in the
beginning of the questionnaire, which may indicate the
identification of the respondent. The code mark is put in a
prominent place in the first page of the questionnaire, like in
the right-hand-top corner of the page. The code may be of
any nature – alfa or numeric or both. However, care should
be taken to see that these codes can be entered into the data
sheet of the SPSS or any other software package, which will
eventually be used for data processing and analysis. In this
code itself the time, date and place of the interview may also
be recorded. However, separate code may be used for this
purpose, or otherwise this may directly be written at in a pre specified prominent place of the first page
of the form.
Request for Cooperation: A request is made to the prospective respondents for their help and
cooperation by the researcher. In this components the researcher should spell out very briefly the
objectives of the research, how this is going to help him or an organisation and why and how the
respondent is selected for the interview. However, sometimes for the sake of extracting unbiased
responses, the identity of the sponsors may not be revealed anywhere in the questionnaire. Many
researcher use creative ideas here to motivate the subject (subject is a world frequently used in place of
respondent in areas like psychology and medicine, where the subjects are monitored for physical or
psychological responses after a particular experiment etc.) to ensure willing and accurate responses
Research Methods in Business
Class Notes: Tezpur University MBA Programme
15
© Mrinmoy K Sarma
from them. The mention of tentative time required to fill up the questionnaire just after the request is of
utmost importance. The required time may be gathered during the pilot survey itself. The average time
taken to fill up the questionnaire during the pilot survey may be used as the tentative time required.
However, if there are many changed after the pilot survey, a new average may be gathered through
another pilot survey.
Instructions are the comments and hints about the questionnaire itself or about individual
questions there in. The instructions help the respondents or the interviewers in fully understanding the
questionnaire and the individual questions and thus help in having accurate measurements of the
variable under study. The instructions common for all questions may be put in the beginning of the
questionnaire just after the request for cooperation. However, the hints regarding a particular question
may be put just after the question in different distinguishable letter font and within brackets. The letter
fonts must be same for all kinds of instructions throughout the questionnaire, and must not be more eye
catching than the letter fonts used for writing the questions itself. Many researcher use italics and one
size smaller (than the size used for the questions) of the same letter font as in the questions. The
instructions should be precise and easily understood by the respondents, and as such this is no place to
show the researcher’s proficiency or knowledge (of word stock) in English. The sentences should be
short and preferably without any idioms.
Information sought is the most important part of the questionnaire, which deals with the
questions itself. This problem is taken up below under the headline Questionnaire Design.
Lastly, classification data are used to classify the respondents on the basis of some predefined
criteria. Income level, age, sex, educational background, profession etc. of the respondents may consist
the classification data. The questions regarding these should normally be asked towards the end of the
questionnaire.
Questionnaire Design
According to Kinnear and Taylor, two well-known authorities in marketing research,
questionnaire design is more of an art than a science. No amount of steps and procedures will ensure
an effective and efficient questionnaire. The skill through which researchers make effective
questionnaire can be acquired only through experience and hard work. The only way to begin is to
develop as many questionnaires as possible and then analyse them for weaknesses and pitfalls.
As mentioned earlier, the measurement of the variables is the very preliminary step towards
writing a questionnaire. Likewise Research Design, Sources of Data, Target Respondents etc. are to be
considered initially before starting the formal process of questionnaire design. These are known as
previous decisions. It must be mentioned here that more heterogeneous the target group more
difficult is the job of preparation of a single questionnaire for the entire group. A perfect link between
the information need and the data to be collected must be established before proceeding. The data
to be collected must have absolute link with the information need. Otherwise the data collected would
not be able to meet the requirement of the research objectives and thus all efforts will be invalid. To
ensure a perfect link, the research objective must be divided into certain sub-objectives beforehand.
The sub objectives then should be tested for the needed information, how the information are to be
collected, what kind of sources of data are to be adopted etc. If primary data were decided to be
collected, the target group of information providers should also be identified. Then the researcher must
decide about the variables those are to be measured to achieve the required sub-objective(s). The
variables are then measured as mentioned earlier.
Research Methods in Business
Class Notes: Tezpur University MBA Programme
16
© Mrinmoy K Sarma
Preliminary Considerations for Questionnaire Preparation
RESEARCH OBJECTIVE (S)
INFORMATION NEEDS
INFORMATION NEEDS
VARIABLES TO BE MEASURED
SOURCES OF DATA
METHODS OF ANALYSIS TO BE USED
TARGET GROUP OF RESPONDENTS
SCALES OF MEASUREMENT
QUESTIONNAIRE
The Scale of Measurement of each variable is to be determined ahead of the beginning of
questionnaire writing. The diagram titled Preliminary Considerations for Questionnaire Preparation
explains the process of arriving at preliminary considerations. [The diagram will be discussed in the
class with examples.]
In the following paragraphs a six-step process of questionnaire preparation is discussed. The
process is depicted in the following diagram.
Steps in Questionnaire Design
Question contents
Response format
Question wording
Question sequence
Physical characteristics
Pilot survey, revise and finalise
Decision about Question Contents
Decision concerning question content centre on the general nature of the question and the
information it is designed to produce according to the preliminary considerations. Five major issues and
problem areas are involved with question contents.
1.
The need for the data asked for by the question: In general, every
question in the questionnaire must be able to make contribution to the problem at hand. Here decision
is to be taken on how the researcher is going to use the data generated by the question? If a
satisfactory answer cannot be provided, the question should not be retained in the questionnaire. A
Research Methods in Business
Class Notes: Tezpur University MBA Programme
17
© Mrinmoy K Sarma
question must be asked if and only if it helps in measuring a particular predetermined variable(s).
However, in case of certain situations an irrelevant question may be asked if the researcher thinks it as
suitable in creating easiness in answering few questions that follows. Or for involvement of the
respondents and creating a rapport before asking a sensitive question. Like if the intention of the
researcher is to ask for TV watching habit of the respondents, he can ask a question whether the
respondent enjoy watching TV.
2.
Ability of the Question in Producing Relevant Data: Once it is assumed
that the question is necessary, the researcher must make sure that the question is sufficient to produce
the required data. Sometimes, a single question may not be enough to measure a particular variable.
Like determining the level of disposable income of the respondent. In this case the researcher may ask
for many indirect or cross-questions at different places of the questionnaire so as to ascertain the level
of the respondent’s spending during a particular period of time, if in the opinion of the researcher a
direct question is not going to give a accurate result. Actually, the researcher may decide about this
while deriving the sub variables for measurement.
3.
Ability of the Respondent to Answer Accurately: Once it is decided that
the question is necessary and sufficient, next task in front of the researcher is to ensure that the subject
knows the accurate answer of the question. Inability in the part of the respondent to answer a question
accurately may be resulted from three situations.
The respondent may have never been exposed to the answer. It is found that the
respondents tend to answer a question even if he/she does not know the answer. This often leads to
serious measurement errors. To avoid this error the researcher may take a two pronged strategy: a)
eliminate such respondents (whom the researcher thinks may not have an accurate answer) in the
sampling frame itself; and b) more realistically, encourage the respondents to leave the questions blank
when they do not know the accurate answer. [Can you find out some tactics for implementing the second strategy?
Also can you visualise what kind of errors might arise if we implement the first strategy wrongly? If you have the answers, offer
them in the class].
The second situation arises when the respondent is forgetful. People are asked
questions, answer to which they once knew but now forgotten. Like the date on which they bought their
TV set or the Microwave oven? Researches have found out that people tend to forget the information
rapidly just after being exposed to them, and then they continuously do so over the passage of time. It is
a well-known fact that the probability of forgetting an event is related to the importance the subject has
attached to the event and the frequency of occurrence of the event. Like if you ask a 50-year-old about
the amount of his first pay cheque, in all probability he will remember it accurately. (Now, whether he will
be willing to share that information with you is another matter.) There are few dangers for the
researcher from the forgetful respondent. Researches have also found out that there might be omission
(simple forgetfulness), telescoping; i.e., the respondent is remembering an event as occurred recently,
while actually it happened long back; and creation, which occurs when the respondent tries to create an
imaginary event.
When the researcher is interested in finding out the facts about an infrequent or
unimportant event the questionnaire designing becomes difficult. (Is not it difficult to define what is
“important” or “unimportant” in the lives of so many respondents you have not met yet). Aided Recall
can be used as one of the solutions, where respondents are given some probable solutions (like in the
multiple-choice questions) of the question. However, in aided recall method, there is a high chance of
occurrence of creation error, which may be reduced by using some bogus or wrong choices in the
solution panel.
The third situation arises when the respondents are able to answer accurately but not
willing to give the accurate answer. There might be two situations again. a) The respondent refuses
to answer. In this case apart from having a high non-response rate, no other error occurs. b) The
respondent willingly offers incorrect answer. There might be many reasons because of which the
respondents may not be willing to cooperate. He may think that the situation is not conducive to offer
the right answer, or disclosure of the data will be embarrassing, or the disclosure might be a potential
threat to the respondent’s prestige or normative view.
Research Methods in Business
Class Notes: Tezpur University MBA Programme
18
© Mrinmoy K Sarma
Here the researcher's job is to motivate the respondents to offer the accurate
response. Financial incentives are widely used to achieve this. However, some other measures like
designing the questionnaire in innocent manner, which assures that the answers are not going to be
used against the respondents may also be used.
One interesting technique, though with some limitations in analysis, is randomised
response technique. This may be used to address the problem of non-response to an embarrassing
question. In this method the researcher presents the respondents with two questions, separately from
the other questions in the questionnaire. One of the two questions will be the intended question, while
the other will be irrelevant for the research and will be very general, like the month of birth of the
respondent, or that of his/her spouse, which the respondent will find very easy to answer. Then a
random procedure, like tossing a coin will be used to select the question to be answered. The question
will be selected by the respondent and the researcher will not know about the selection. Essentially, the
response format should be same for both the questions. And the respondent will tell the interviewer only
the answer, like yes or no. Since the respondents are assured that the interviewer will not know the
question he/she is answering, it is expected that the respondent will behave reasonably and offer the
accurate answer to whatever question the random procedure selects. After completion of the data
collection procedure the researcher use the help of some secondary data to find out the proportion of
the respondents said yes or no (or whatever the response format was) to the question under
consideration.(Let us play a game in the Class)
An example of Randomised response Technique: Suppose you are interested in finding
out the percentage of college going students who drink. If you ask a teenager directly about whether
he/she drinks chances of getting a biased (incorrect) answer are very high. To overcome this you may
club this question with a very easy question like the month of his/her birth. You must be careful in
selection of the second question as this should meet the following two criteria:
The question should be very easy to answer
And the answer to the question should be available from a secondary source.
Now consider the following example of a randomised response questionnaire:
TOSS THE COIN THE INTERVIEWR IS GIVING YOU.
IF YOU GET HEAD, ANSEWR QUESTION NO.1 BELOW.
IF TAIL APPEARS ANSWER TO QUESTION NO. 2.
DO NOT TELL OR SHOW THE INTERVIWER ANYTHING.
THUS INTERVIWER WILL NOT BE KNOWING TO WHICH QUESTION YOU ARE ANSWERING.
AND YOUR PRIVACY WILL REMAIN INTACT
1.
YOU WERE BORN IN THE MONTH OF DECEMBER.
2.
YOU HAVE NEVER STOLEN ANY MONEY FROM YOUR FATHER’S PURSE.
ANSWER:
YES
NO
Answer to the first question will be known from the data collection by the census
authority (do you know the name of the office which conduct census in India?). Even from intuition the
answer can be guessed to be around
1
. Now the percentage of respondents which answers “No” to
12
the second question could be calculated by the following formula.
P( No)  P(insensitiv e _ question) * P(" No" to _ insensitiv e _ question)
P( sensitive _ question)

See Thomas C Kinnear and James R Taylor ,“Marketing Research”, 5 th international edition, McGraw-Hill for
more detail.
Research Methods in Business
Class Notes: Tezpur University MBA Programme
19
© Mrinmoy K Sarma
Where,
P (No) is proportion of respondents who answered "No"
P (insensitive_question) is the probability of answering 'No" to the insensitive question (in this
case it is 0.5)
P ("No" to_insensitive_question) is proportion of people born NOT in December (from census
data)
P (sensitive_question) is the probability of answering 'No" to the sensitive question (in this case
it is 0.5)
Decision About Response Format
This is one of the more creative areas of questionnaire preparation. Normally there are two
types of response format. 1. Open ended question, and 2. Close ended questions. There are various
advantages and disadvantages of both the types of questions.
To make the questionnaire attractive multiple choice questions may be used in many ways.
Interesting pictures, cartoons etc. may provided in lieu of normal words. Like smiley faces might be
used for expressing satisfaction level. The following example will give an idea.
Question: Are you satisfied with the after sales service of the product XYZ?
Choices:
Very
Satisfied
Somewhat
Satisfied
Indifferent
(cannot say)
Not
Satisfied
Not
Satisfied at all
These choices
can graphically
be depicted as :
Decision on Question Wording:
The heart of the questionnaire consists of the questions - the link between the data and
information needs of the study. It is critical that the researcher and the respondents assign same
meaning to the questions asked. Otherwise serious measurement errors will occur. Therefore, the
discussion under this sub-head will revolve around how to write questions, which carry the same
meaning for both researcher and the respondent. There are certain general guidelines regarding
designing the wordings of the question.
1.
Use simple words
2.
Use clear words: In finding out clarity of words used in a questionnaire,
answer the following questions.
a) Does this word mean what is intended?
b) Does it have any other meaning?
c) If so, does the context make the intended meaning clear?
d) Does this word have more than one pronunciation? (for Telephonic interview this is
most important)
e) Is there any word with similar pronunciation that might be confused with this word?
f) Is there a still simpler word or phrase that might be used?
1.
Avoid leading questions: A leading question is one in which the respondent is
given a cue as to what should be the answer of the question. Leading question often
reflects the researcher's or the organisation’s view point on a particular variable under
study. This type of question causes constant measurement error throughout the survey.
An example of leading question would be to ask a respondent: Today is very hot. Is not
it?
Research Methods in Business
Class Notes: Tezpur University MBA Programme
2.
20
© Mrinmoy K Sarma
Avoid biasing questions: A biasing question includes word or phrases those
are emotionally coloured and suggest a feeling of approval or disapproval. Often it
includes certain references, which might effect the answer of a question. An example
might be Do not you agree with the view of Bill gates that new century will be an era of
Information Technology? And who would, in earth, like to answer such kind of an
question? Those who like Bill will say yes and those who do not know gates will say no.
And now your job is to find out which one is the honest response!
An example of loaded question will be Is General Motors doing everything
possible in reducing automobile pollution? The answer is obvious no or obvious yes,
because General Motors is doing, and also not doing everything possible to reduce
automobile pollution. Even if the intention of the researcher was different the use of the
two words made the question vulnerable to measurement error.
3.
Avoid implicit alternatives: Researcher should not offer only partial
alternatives in case of close ended questions. If your job is to find out the brand name
of the Toothpaste used by the respondent, then you must include all available brands in
the market so that while responding the sample does not feel difficulty in finding out a
place to tick. If it is not possible to include all the alternatives or if your job is not to find
out the exact brand of the toothpaste, you are allowed to use the word others so that
any measurement error does not occur. That means the choices should be collectively
exhaustive.
4.
Implicit assumption: Consider the two questions: Do you favour a ban on
commercials in movie theatres? And just add the following part to the question. …even
if it means a rise in $.5 per show? The proportion of samples that said ‘yes’ to the first
part of the question is 22%, and the same for the whole question is 11%. Often it is
found that measurement error galore for the failure of the researchers in stating
essential assumptions. Therefore, the researcher must not think that the respondents
know the virtual assumptions related to a question.
5.
Avoid estimates : Question should not be such that the respondent has to
rely on estimates. Therefore, distant recall question should not be asked. If estimate is
encouraged the respondents might give inaccurate information.
6.
Avoid double barreled questions: The questions should be simple and
should contain only one answer. However, it is often seen that one question actually
includes two questions and thus two answers. Consider, What is comment on LML
Vespa's after sales service and economy standard? The answer to these two questions
may be different, and the respondent is not sure to which question you are asking
answer!
7.
Consider the frame of reference: This refers to the respondent's view point
from which he/she is answering the questions. Consider the two questions: Are
automobile manufacturers are making satisfactory progress in controlling automobile
emission? And Are you satisfied with the progress automobile manufacturers are
making in controlling emissions?
The wording for each question should be as simple as possible. It should be consistent to the
vocabulary level of the respondents.
Decide on Question Sequence: once the wordings are finalised, the questions should be put in
sequence. The following guidelines may be followed for this purpose.
Use simple and interesting opening question
Ask general questions first
Place uninteresting question later in the sequence.

S. Hume and M. Magiera, "What do Moviegoers Think of Ads?" Advertising Age ( April 1990)
Research Methods in Business
Class Notes: Tezpur University MBA Programme
21
© Mrinmoy K Sarma
Arrange questions in logical order.
Physical get up of the questionnaire should be eye-catching, gorgeous and should be full of
blank spaces. The letters should be easily legible to all members of the sample.
Pilot Survey: The basic job of a pilot survey is to find out the difficulties that might be faced
during the field operations of the survey with the questionnaire so that necessary actions and
modifications can be initiated to remove these difficulties in the time of actual field operation. Specially
the understanding of the questions by the respondents are tested. The following issues are taken up in
this stage:
In terms of physical appearance:
a)
Will the questionnaire appeal to respondents and motivate them to cooperate? Is
your questionnaire "sensuous"?
b)
Does the questionnaire include brief and precise instructions? Are the
instructions are enough to explain the respondents what is wanted from them?
Are they confusing in any respect?
c)
Is your format conducive to your chosen method of data entry? (like keying,
scanning and hand tabulation etc?
In terms of content:
a)
Does each question ask for one bit of information?
b)
Does the question presuppose a certain state of affairs? If so, if the assumption
is justified?
c)
Does the question wording bias response?
d)
Are any of the questions words emotionally loaded, vaguely defined, or overly
general?
e)
Do any of the question's words have a double meaning, which may confuse
respondents?
f)
Does the question use abbreviations, or jargon, which may be unfamiliar?
g)
Are the question's responses mutually exclusive and sufficient to cover each
conceivable answer?
Research Methods in Business
Class Notes: Tezpur University MBA Programme
22
© Mrinmoy K Sarma
SURVEY METHOD
There are three widely used methods of primary data collection, especially through survey method.
These are,
 mail,
 telephone interview, and
 in-person interview
There remain inherent merits and demerits of all these methods. The methods are evaluated
on the basis of the following criteria







Versatility of use
Cost associated
Time taken
Possibility of Sample Control
Quality of information
Quantify of information
Response rate (or non response rate)
Mail surveys are the most common examples of self-reported data collection. One of the
reasons is that such surveys can be relatively low in cost. This does not mean, however, they are
necessarily easy to carry out. Planning the questionnaires for mail surveys is often more difficult than for
surveys those use interviewers (in-person interview or telephonic interviews). For example, care is
needed to anticipate the physical and psychological conditions of the respondents in advance – in many
cases without knowing the ground realities of the environment the respondent is facing.
Using the mail (post) can be particularly effective in business surveys. Mail surveys also work
well when they are directed toward specific groups -- such as, subscribers to a specialized magazine or
members of a professional organization. The manner in which self-reported data are obtained has
begun to move away from the traditional mail-out/mail-back approach. The use of fax machines -- and
now the Internet -- is on the rise. Fax numbers and e-mail addresses are being added to specialized
membership and other lists. As a by-product, they can be used, along with more conventional items like
names and mailing addresses, in building potential sampling frames.
There are still other methods of obtaining self-reported data like the one used for obtaining
continuous information about the same elements over a period of time. Panels are such examples.
However, technology has helped reducing the cost, time and effort in collecting such routine
information. For example, computer network can be used to put in necessary information into the
principal server by the remote respondents. However, for the immediate future, this type of automation
will probably be restricted largely to business or institutional surveys in which the same information is
collected at periodic intervals -- monthly, quarterly, etc. Do you think TRP surveys can use this
technique?
How to Conduct an Interview
Interview surveys -- whether face-to-face (in-person) or by telephone -- offer distinct
advantages over self-reported data collection. The "presence" of an interviewer can increase
cooperation rates and make it possible for respondents to get immediate clarifications. One of the main
requirements for good interviewers is the ability to approach strangers in person or on the telephone
and persuade them to participate in the survey. Once a respondent's cooperation is acquired, the
interviewers must maintain it, while collecting the needed data, which should be obtained in exact
accordance with laid down instructions.
For ensuring quality of the collected data, interviewers should be carefully trained through
classroom instruction, self-study, or both. Good interviewer techniques such as -- how to make initial
contacts, how to conduct interviews in a professional manner, and how to avoid influencing or biasing
responses. Training generally involves practice interviews to familiarize the interviewers with the variety
of situations they are likely to encounter. However, for different interviews, the interviewers should be
Research Methods in Business
Class Notes: Tezpur University MBA Programme
23
© Mrinmoy K Sarma
trained separately so that a question by question understanding is achieved to make them qualified to
deal with any misunderstanding that might arise at the time of the interview. Also, the interviewers may
be made clear about the purpose, definitions and procedures of conducting a particular survey
separately. In most reputable survey organizations, the interviewers are also required to take a strict
oath of confidentiality before beginning the work. Interviewers should also be trained on the way the
samples are to be selected, if needed. If they are to visit the pre-selected samples, adequate guiding
materials such as addresses, maps, pictures etc. should be made available to them (after imparting
training on how to read them), so that they make no mistake in finding the right samples.
It is advisable to send an advance letter to the sample respondents, explaining the purpose of
the survey and that an interviewer will be calling soon. Many reputed survey organisations offer
information to the respondents on how the information will be used and the level of confidentiality of the
data.
Visits to sample units should be scheduled with attention to considerations like the best time of
day to call or visit, which might be gathered from the pre-selected samples through advance call or mail.
Computers and Survey:
Few years back more than 95% of the 165 members of the Council of American Survey
Research Organisation offered internet based data collection, wherein they used internet as their data
collection tool rather than typical meet-the-respondent fill-up-the-questionnaire technique. They reported
many advantages of such a method of data collection. Speed is one of the main advantages. One
market research organisation completed 1000 questionnaire for a customer satisfaction survey within
only 2 hours! This is incredible compared to the time (sometimes months) required in traditional
method. Networked research also offers the ability to target hard-to-reach population. One of the
traditional difficulties in segmentation research is to identify and access respondents who fit a particular
lifestyles or reside in a remote area. Now-a-days many internet portals offer segmentation statistics at a
price and thus the researchers can reach such population without much difficulties. You might have
noticed while using internet how the promotional mails are sent (most of the cases unsolicited). The
same method can be used with a bit of refinement and modification. Web based data collectors can use
the opportunity for multi-media presentation to make their points. With the advent of high-speed
network connection (in giga-bites) this would be more practical and user friendly. And remember this
can be made at the disposal of the respondents without any movement of the interviewers. However,
despite all these unparalleled advantages such research may be infected with the traditional errors of
data collection. More so because of the fact that the Internet addresses are impossible to verify to
ascertain whether the sampling frame is the correct one.
However a blending of traditional method and computer method can reduce these errors to the
minimum, while keeping the distinctive advantages of the networked survey.
Computer Assisted Telephone Interview (CATI): The use of computers in survey
interviewing is increasing day by day. American Statistical Association reports that in the United States,
most of the large-scale telephone surveys are now conducted using computers. In CATI, the
interviewers use a computer either in a network or stand alone to conduct the telephonic interview. The
questions in order of preference appear in the screen and the responses are inserted directly into the
computer. Then the same are analysed readily using the required statistical software.
The CATI interviewer's screen is programmed to show questions in a planned order, so that
interviewers cannot inadvertently omit questions or ask them out of sequence. Specially, if some
questions require “branching” (i.e., answers to prior questions determine which other questions are to
be asked. Like, if the answer to a question is “yes”, then a different set of questions are asked, while for
“no” as an answer still different questions are asked) CATI can be programmed to do the correct
branching automatically. In ordinary telephone interviewing, incorrect branching has sometimes been an
important source of errors, especially omissions. CATI can also be used to make automatic crosschecking of responses. If certain inconsistency occurs, the software itself will pop in certain question on
the screen for the respondent to clarify (to correct or confirm) earlier responses.
Research Methods in Business
Class Notes: Tezpur University MBA Programme
24
© Mrinmoy K Sarma
Other advantages of this method of telephonic interview are quality and speed. CATI can
produce statistical results quicker than traditional methods of data collection. For example, it eliminates
the need for a separate data processing and data entry. This method is more useful when a daily or
periodic summary of results is required.
However, limitations of CATI include the type of questions – obviously only close ended
questions with multiple choices can be managed through such methods. Any insertion, which requires
longer time, may distract the respondent, as the waiting time for the next question will be long.
Moreover, CATI can cost more for small, non-repeated surveys, due to programming the questionnaire.
CATI's cost per interview decreases as sample size increases -- so in large and/or repeated surveys, it
is cost competitive with conventional telephone methods.
Computer Assisted Personal Interview (CAPI): This method has direct linkage to the high
level of use of lap top computers or other portable computer systems, which can be taken into the field,
and either the interviewer or the respondent can directly enter data in response to questions. Data
collection carried out in this way is referred to as CAPI. The CAPI laptops may not be in a network at
the time of administration of the questionnaire. Nonetheless, most CATI quality and speed advantages
also occur with CAPI.
Although only a few organizations currently employ CAPI methods, their use is expected to
expand in the next few years. Clearly, the periodic interviews like that of a panel study (for example to
determine TRP {do you know by this time what is it? If not, find out NOW} indexes) may be greatly
benefited by the use of CAPI.
However, the question remains as to what extent the traditional paper and pencil method will
remain as the prime tool of conducting interviews! Who can predict!!
Shortcuts to Avoid during a Survey:
A credible survey must be carefully planned and controlled (during execution). This needs lots of
determination, consistency in the approach and perseverance. Amateur researchers are often inclined
to adopt shortcuts, as they feel such measures would not jeopardize with the quality of the findings.
However, contrary to their belief, taking shortcuts can invalidate the results and badly mislead the
sponsor and other users. Here three most commonly used shortcuts are mentioned.



Pretesting of field procedures (pilot survey) is avoided
non-respondents are not followed up sufficiently
Sloppy fieldwork and inadequate quality controls
Therefore, efforts should be made not to take these shortcuts for the sake of collecting good
quality data. When non-response occurs, efforts must be made to re-contact the sample again and
again. Every organisation might have some policy as to how the non-responses are taken care of, or
how many times a non-respondent will be tried to be contacted. However, if non-response is occurring
for reasons other than non availability of the sample at the time of the visit of the interviewers or a
returned mail due to non availability of the addressee, non-response can be prevented to a great extent
by proper planning and pretesting of the questionnaire. A pretest of the questionnaire and field
procedures is the only way of finding out if everything “works”— especially if a survey employs new
techniques or a new set of questions. We will discuss about Pretesting, which is also known as pilot
survey is a later part of this material. Sloppy execution of a survey in the field can seriously damage
results, Controlling the quality of the fieldwork is done in several ways, most often through observation
or redoing a small sample of interviews by supervisory or senior personnel. There should be at least
some questionnaire-by-questionnaire checking, while the survey is being carried out; this is essential if
omissions or other obvious mistakes in the data are to be uncovered before it is too late to fix them. In
other words, to assure that the proper execution of a survey corresponds to its design, every facet of a
survey must be looked at during implementation. For example... re-examining the sample selection …
redoing some of the interviews. Without proper checking, errors may go undetected. With good
procedures, on the other hand, they might even have been prevented. As W. Edwards Deming
recommends, a complete systems approach should be developed to be sure that each step fits into
the previous and subsequent steps. Murphy’s Law applies here, as elsewhere in life. The corollary to
Research Methods in Business
Class Notes: Tezpur University MBA Programme
25
© Mrinmoy K Sarma
keep in mind is that not only it is true that “If anything can go wrong it will… but, “If you didn’t check on
it, it did.”
How to Plan a Survey:
To begin with every researcher must ask the following questions repeatedly:
1.
2.
3.
Whether the required information can be collected through a survey?
Or may be these cannot be?
Is the information available in some indirect sources?
When the researcher is satisfied with the answers and convinced that there is a need for survey
for the required information, he/she can take further steps in planning a survey. The following stages of
activities are generally followed while planning a survey.
A.
B.
C.
D.
E.
F.
G.
H.
LAY DOWN THE OBJECTIVES OF THE INVESTIGATION: This is generally is
the function of the sponsor of the survey. However, it is the duty of the
researcher to finalise the objectives which are achievable (means, not
unrealistic) with the consultation of the sponsor. The objectives of the survey
should be as specific, clear cut and unambiguous as possible.
SPECIFY THE DATA COLLECTION PROCEDURE: The mode of data
collection must be decided upon before proceeding further. The decision will
have to be made whether the mail (conventional or electronic), telephone or inperson method will be applied. The steps those follow this will be heavily
dependent upon this decision of the researcher.
PLANNING OF THE DATA COLLECTION FORM: If the researcher is willing to
employ the mail in and mail out method of data collection, the form of data
collection, which is known as the questionnaire, in this case will have to be
carefully planed and implemented (please see the topic questionnaire, which s
included in the material.). However, if the electronic method is used proper
care has to be taken to see that enough responses are received with out any
distortion of the objectives. Like if the researcher is administering the
questionnaire through e-mail, the planning will be different than if the data are
to be gathered through submissions in the web page. This stage also does
have a bearing with the target group or the population from which the data are
to the extracted. This factor leads to the next stage of the planning process.
DECIDING ABOUT THE POPULATION OR THE SAMPLING FRAME: The
group of persons from whom data are intended to be gathered is known as the
population of the survey. When each and every member of the population is
featured in a comprehensive list, this is known as the sampling frame.
However, in many of the social science researches, the sampling frame is not
available (see sampling frame for detail). However, it must be remembered that
for conducting a scientific survey with a probabilistic method of data collection
(like simple random sampling, stratified random sampling or area sampling) the
presence of a sampling frame is must.
THE DATA COLLECTION PROCEDURE: The next step is to decide about the
data collection procedure. This is discussed in detail elsewhere in this material.
DECIDE ABOUT THE SCHEDULE OF THE SURVEY: This is one of the
important decisions the researcher will have to take before starting of the
survey. Because the quality of data is totally dependent on the time schedule of
the survey. In certain cases like that of an opinion poll or exit poll, or test
marketing of a product the time factor is very important and precise planning
must be made to follow the desired time schedule.
SELECTION OF SAMPLE INCLUDING DECIDING ABOUT THE SAMPLE
SIZE: These are discussed elsewhere in this material.
ADMINISTRATION OF THE DATA COLLECTION FORM: This is the fieldwork
when the interviewers actually go down to meet the sample. If this is not an in-
Research Methods in Business
Class Notes: Tezpur University MBA Programme
26
© Mrinmoy K Sarma
person data collection survey, this step includes mailing out the questionnaire,
or making phone calls to the samples.
The Budget for a survey: The monetary involvement of a survey is a function of the
data collection procedure and the sample size. Following is a checklist for allocation of cost as
prescribed by the American Statistical Association. A “traditional” (paper and pencil) in-person interview
survey will be used to illustrate the budget steps. Many of these are general; however, increasing use of
survey automation is altering costs — reducing some and adding others.












Staff time for planning the study and steering it through the various stages,
including time spent with the sponsor in refining data needs.
Sample selection costs, including central office staff labour and computing costs.
For “area segments” samples, substantial field staff (interviewer) labours costs and
travel expenses for listing sample units within the segments.
Labour and material costs for pretesting the questionnaire and field procedures; the
pretesting step may need to be done more than once and money and time should
be set aside for this (especially when studying something new). Supervisory costs
for interviewer hiring, training, and monitoring.
Interviewer labour costs and travel expenses (including meals and lodging, if out of
town).
Labour and other costs of redoing a certain percentage of the interviews (as a
quality assurance step) and for follow-up on non-respondents.
Labor and material costs for getting the information from the questionnaire onto a
computer file.
Cost of spot-checking the quality of the process of computerizing the paper
questionnaires.
Cost of “cleaning” the final data— that is, checking the computer files for
inconsistent or impossible answers; this may also include the costs of “filling in” or
imputing any missing information.
Analyst costs for preparing tabulations and special analyses of the data; computer
time for the various tabulations and analyses.
Labor time and material costs for substantive analyses of the data and report
preparation.
Potentially important are incidental telephone charges, postage, reproduction and
printing costs for all stages of the survey — from planning activities to the
distribution of results.
A good survey does not come “cheap,” although some are more economical than others
are.
Research Methods in Business
Class Notes: Tezpur University MBA Programme
27
© Mrinmoy K Sarma
DATA ANALYSIS
Data analysis is another very important step of the research process. In this stage the collected
data are analysed by the desired and appropriate method.
Data are of different types depending on the number of variables taken for analysis. When
there is only one variable in consideration than this is known as Univariate data. When the number
increases to 2 then it is known as Bivariate data and when the variable in consideration is more than 2
the data are known as Multivariate data. Let us discuss the available methods of Univariate data.
Before proceeding further let us have a discussion regarding Hypothesis Testing.
HYPOTHESIS TESTING:
We must state the assumed value of the population before we begin sampling. The assumption
we wish to test is known as null hypothesis. This is symbolised by H0 If we are to test the hypothesis
that the population mean is 500 then this is to be written as
Ho : = 500
In null hypothesis, generally a situation is tested which does not have any effect i.e., there is no
difference between treated and untreated samples. Suppose the management of a firm believes that
the average consumption of carbonated beverage per female student per week was more than four
bottles. If the consumption of the beverage is this high the co. may develop a different kind of drink to
suit the consumer in a different way. But, before proceeding the co. must be very sure about the
consumption level. However, after the sample analysis they found that the av. consumption level is
5.575. Even though for further surety the co. formulated the Null Hypo. as
Ho :  < 4 bottles of beverages per week.
Here, the firm formulated the hypothesis in the null form, i.e., the av. consumption per female is
less than 4 bottles.
There is another type of hypothesis known as H1 or alternative hypothesis. Here the alternative
hypothesis may be
H1 : > 4 bottles.
The purpose of hypothesis testing is not to question the computed value of the sample statistic
but to make a judgment about the difference between that sample statistic and a hypothesized
population parameter.
STEPS IN HYPOTHESIS TESTING:
1.
2.
3.
4.
5.
6.
Formulate null and alternative hypotheses.
Select the appropriate statistical test given the type of data the researcher has.
Specify the significance level.
Look up for the value of the test statistics in a set of table for a given level of significance
Perform the statistical test. This yields a value of the relevant statistic.
Compare the value of the statistics calculated in item 5 with the value of the item 4. If the value
of item 5 is greater than the value of item 4, then reject the null hypothesis.
The steps described above are for standardised hypothesis testing method. In Traditional
Method the data thus gathered are plotted in a normal curve and then the critical value from the table is
Research Methods in Business
Class Notes: Tezpur University MBA Programme
28
© Mrinmoy K Sarma
found out. If the calculated value of the statistic falls in the acceptance region than the null hypothesis is
accepted and otherwise rejected. (These are discussed in the class in detail).
Type one error occurs when a correct null hypothesis is rejected.
Type two error occurs when a wrong null hypothesis is accepted.
Significance level indicates the probability of type one errors being made. In fact this is the tolerance
level of the researcher of type one error.
Two tailed and one tailed test of hypothesis:
A two tailed test of a hypothesis will reject null hypothesis if the sample mean is significantly
higher than or lower than the hypothesized population mean. Thus In a two tailed test there are two
rejection zones.
A two tailed test is appropriate when the null hypothesis is  = H0 ( H0 being some specified
value) and the alternative hypothesis is   H0. Assume that a manufacturer of light bulbs wants to
produce bulbs with mean life of  = H0 = 1000 hours. If the lifetime is shorter, he will loose customers to
his competitors; if lifetime is longer he will have to bear extra cost in production. Since he does not want
to deviate significantly in either way, a two-tailed test will be appropriate.
However, if a wholesaler that buys bulbs from the above manufacturer, he would not accept
bulbs with life less than 1000 hrs. However, he is not bothered if the life span of the bulbs are higher
than 1000hrs. In this case the wholesaler will be interested in testing whether the bulbs are less than
1000 hrs of life. Hence his null hypothesis will be H0: <1000 hours. In this case one tailed test (left
tailed) is appropriate.
The z test:
Appropriate situation for using z test:
1. the sample of any size if the population standard deviation is known
2. the sample size is greater than 30 if population s.d. is not known.
When n 30 and population s.d. is not known t test is appropriate.
If  is known
z
x x
=

x
n
If  is NOT known
z
x x
=
s
sx
n
Research Methods in Business
Class Notes: Tezpur University MBA Programme
29
© Mrinmoy K Sarma
The following flow chart gives an idea regarding use of different types of univariate data
analysis techniques. It is clear from the first question in the flow chart that the decision regarding the
tools to be used would primarily depend upon the scales of measurement used. Therefore, if you
contemplate to use a particular tool (even in case of bivariate and multivariate techniques) chose the
scale accordingly while specifying the instruments of data collection. And obviously this is to be done
before collection of data.
Univariate data analysis procedures:
What is the scale level of the variable ?
interval
ordinal
nominal
1. Descriptive
Central
Tendency
Mean
Median
Mode
Dispersion
Standard
Deviation
Interquartile
Range
Relative & absolute
frequencies by
categories
z test
t test
Kolmogorov
Smirnov test
Chi-square test
2. Inferential